Patient and Provider Factors Associated With American Indian and Alaska Native Adolescent Tobacco Use Screening

Abstract

Introduction: Tobacco use is the leading behavioral cause of death among adults 25 years or older. American Indian (AI) and Alaska Native (AN) communities confront some of the highest rates of tobacco use and of its sequelae. Primary care–based screening of adolescents is an integral step in the reduction of tobacco use, yet remains virtually unstudied. We examined whether delivery of tobacco screening in primary care visits is associated with patient and provider characteristics among AI/AN adolescents. Methods: We used a cross-sectional analysis to examine tobacco screening among 4757 adolescent AI/AN patients served by 56 primary care providers at a large tribally managed health system between October 1, 2011 and May 31, 2014. Screening prevalence was examined in association with categorical patient characteristics (gender, age, clinic visited, insurance coverage) and provider characteristics (gender, age, tenure) using multilevel logistic regressions with individual provider identity as the nesting variable. Results: Thirty-seven percent of eligible patients were screened. Gender of both providers and patients was associated with screening. Male providers delivered screening more often than female providers (odds ratio [OR] 1.6, 95% confidence interval [CI] 0.7-3.9). Male patients had 20% lower odds of screening receipt (OR 0.8, 95% CI 0.7-0.9) than female patients, independent of patient age and provider characteristics. Individual provider identity significantly contributed to variability in the mixed-effects model (variance component 2.2; 95% CI 1.4-3.4), suggesting individual provider effect. Conclusions: Low tobacco screening delivery by female providers and the low receipt of screening among younger, male patients may identify targets for screening interventions.

Introduction

Tobacco use is the leading cause of preventable death in the United States^1,2 and the leading behavioral cause of death among adults 25 years or older.³ Tobacco use typically begins in early adolescence, with early experimentation progressing to nicotine dependence by age 18 years.^4,5 Roughly a third of youth who become regular tobacco users before adulthood will eventually die from a tobacco related illness.⁶ Negative health consequences are especially pronounced in vulnerable subpopulations, such as racial/ethnic minorities, who have particularly high rates of tobacco use accompanied by elevated rates of tobacco-related cancers and other health conditions.⁷

American Indian (AI) and Alaska Native (AN) communities confront some of the highest rates of tobacco use and of its sequelae,^1,8-11 including rates of lung cancer that are 40% higher than among US whites¹² and earlier mortality.¹³ Within the AI/AN adolescent population, tobacco use typically begins before high school, with progression to regular tobacco use occurring in the early teen years. Because 34% of all AN high school students report smoking a whole cigarette before age 13 years, screening for tobacco use should begin in early adolescence before nicotine dependence occurs.¹⁴ Primary care–based screening, referral for treatment, and successful cessation intervention among AI/AN adolescents is an integral step in the reduction of tobacco use among AI/AN people, yet remains virtually unstudied.

The US Preventive Services Task Force recommends that health care system staff screen patients for tobacco use, advise to quit, and refer users to complete counseling interventions.^15,16 Primary care providers may be the first or only health care professionals individuals encounter who can recognize substance use and intervene as problems develop.¹⁷ Substance use screening and referral to treatment processes are intended to target populations at the earliest stage of use to identify and intervene at the earliest stage of use.¹⁸

However, not all patients are equally likely to receive screening.^19-24 A number of studies, several focused specifically on racial/ethnic minority subpopulations, suggest that patient race and gender may be positively associated with screening for tobacco use.^25-27 Tobacco use among AI/AN rural youth is higher than among their urban counterparts, while access to screening and referral services may be lower in their communities.^28-30 Provider characteristics may also be associated with screening, as in one study of adolescent patients in a health maintenance organization where female pediatricians more often provided preventive services, including tobacco screening, than did their male counterparts.³¹ Still other studies have found that patients with gender-concordant providers were both more likely to be referred for treatment and to have positive health outcomes from treatment than were patients with gender-discordant providers.^32,33

While researchers have urged investigation of the possible influence of patient and provider characteristics on screening as a step toward improving patient-centered care,³⁴ these remain poorly understood in patient subgroups. Only one study has examined tobacco screening in AI/AN adolescents.³⁵ Accordingly, ours is the first large-scale project to examine patient and provider characteristics that may influence tobacco use screening in this racial/ethnic population. Our analyses included variables with known importance for provider-patient interaction in the general population, including patients’ and providers’ gender and gender concordance, patients’ age and rurality, and providers’ years tenure at the screening facility.

Methods

Study Setting and Design

The study setting was Southcentral Foundation (SCF), a nonprofit, tribally owned and operated health care organization. SCF provides a wide range of health and human services to more than 65 000 AI/AN people in the Indian Health Service Anchorage Service Unit via 14 primary care clinics in both urban and remote rural villages.^36,37 Southcentral Foundation’s tobacco screening process in all 14 primary care clinics is to screen all patients aged 12 years and older for tobacco use in the examination room. Patients are verbally asked by a health care provider “Do you use tobacco?”³⁸ Verbal responses question are immediately recorded in the patient’s electronic medical record.

A cross-sectional analysis was used to examine adolescent patients (12-18 years old) who received primary care services at Southcentral Foundation outpatient primary care clinics, outpatient pediatric clinics, or rural clinics affiliated with the Alaska Native Medical Center. The data were extracted from the SCF’s electronic medical records that were implemented clinically on October 1, 2011. Data were extracted by a single abstracter using electronic querying for service utilization dates spanning October 1, 2011 through May 31, 2014.

Study Sample and Measures

The data were organized into a cross-sectional cohort of individual, unique patient-visits, and were analyzed on that scale. Inclusion criteria for the study included the following: (a) patient had AI/AN heritage; (b) patient age at the date of service was between 12 and 18 years; (c) visit occurred with a primary care provider in outpatient primary care, outpatient pediatrics, or rural clinics affiliated with the Alaska Native Medical Center; and (d) visit occurred between October 1, 2011 and May 31, 2014. Exclusion criteria included telephone contacts or visits for blood draws, as these types of visits are not part of the provider protocol for tobacco screening. There were 63 128 unduplicated, eligible patient-visits. For patients who were seen by a provider multiple times during the data extraction period, we selected the first eligible visit resulting in a study cohort of 4757 individual patient-visits meeting the inclusion criteria.

Patient demographic characteristics were based on a previous primary care screening study³⁹ in this health system and data were extracted from the patient charts included gender (male, female), age at eligible visit (12-13, 14-15, 16-17, 18 years), clinic location (rural, urban/Anchorage), and patient insurance coverage (any private insurance, Medicare/Medicaid only, no coverage). Patient health data extracted from individual clinic visit records included evidence for screening for tobacco use (yes, no). Patient health service utilization patterns over the 12 months prior to the baseline visit (none, 1-4, 5-9, 10+ visits) was deduced from the set of eligible clinic visits extracted from the medical records database. Provider characteristics were extracted from the SCF human resources database, including provider gender (male; female), age at clinic visit (18-35, 36-55, 55+ years), and duration of tenure at SCF (<2, 2-5, 5-10, 10-15, 15+ years). Provider race was not included in the data set, preventing analyses of provider-patient racial concordance.

Statistical Analyses

Screening prevalence was calculated and stratified by categorical patient and provider characteristics. Patient and provider characteristic associations with the occurrence of tobacco use screening were evaluated using multilevel logistic regressions with provider identity as the nesting variable. Multilevel models used the conditional distribution of the outcome variable (tobacco screening), given the random effects from the nesting variable (provider ID). Odds ratios (ORs) were calculated with 95% confidence intervals (CIs) to examine the associations between various patient or provider characteristics (exposure) and the occurrence of tobacco screening (outcome), adjusted for potential confounders (patient age, gender; provider age, gender, tenure) and the random effect (individual providers). Multilevel model coefficients are interpretable as individual effect, or the change in the outcome, given a certain change in exposure, holding other model variables constant. Secondary analyses were stratified by patient gender. Similar analyses compared different patient-provider gender dyads in association with tobacco screening, adjusted for patient age, provider age, provider tenure. All analyses were conducted using Stata version 11-13 (StataCorp, College Station, TX). All data collection and analytic procedures were approved by the Indian Health Service Alaska Area Institutional Review Board and internal review by Southcentral Foundation and the Alaska Native Tribal Health Consortium.

Results

The study cohort of individual patient-visits was composed of 4757 AI/AN adolescents seen by 56 primary care providers (Table 1). Thirty-seven percent (37%, n = 1753) of these patients were screened for tobacco use over the course of their clinic visit. A majority of patients were female and older than 16 years, and had more than 5 visits to the SCF health system in the year prior to their index visit. Providers also tended to be female, between 36 and 55 years, and with more than 5 years tenure. Visits occurred mostly in the urban setting (ie, Anchorage or surrounding area), and comprised patient-provider dyads of the same gender.

Table 1.

Tobacco Screening of American Indian and Alaska Native Adolescent Primary Care Patients by Patient Factors.

	Screened (n = 1753)		Not Screened (n = 3004)
	n	%	n	%
Patient factors
Gender
Female	904	51.6	1592	53.0
Male	849	48.4	1412	47.0
Age, years
12-13	107	6.1	167	5.6
14-15	620	35.4	724	24.1
16-17	498	28.4	866	28.8
18	528	30.1	1247	41.5
Residence
Rural (eg, not Anchorage)	342	19.5	1151	38.3
Insurance status
(Any) Private insurance	657	37.5	1142	38.0
Medicare/Medicaid only	782	44.6	1174	39.1
Number of clinic visits in year prior
None	5	0.3	0	0
1-4	316	18.0	867	28.9
5-9	453	25.8	924	30.8
10+	979	55.9	1213	40.4
Other substance use screenings	794	45.3	1064	35.4
Provider factors
Provider gender
Female	1154	65.8	2267	75.5
Male	599	34.2	737	24.5
Provider age, years
<35	233	13.3	904	30.1
36-55	954	54.4	1511	50.3
>55	566	32.3	589	19.6
Provider tenure, years
<2	113	6.5	262	8.7
2-5	517	29.5	964	32.1
>5-10	531	30.3	998	33.2
>10-15	575	32.8	720	24.0
16+ years	17	1.0	60	2.0
Provider-patient dyads
Female provider, female patient	1337	44.5	675	38.5
Female provider, male patient	930	31.0	479	27.3
Male provider, male patient	482	16.1	370	21.1
Male patient, female patient	255	8.5	229	13.1

The multilevel, multivariate logistic regressions examining the occurrence of screening indicated that male patients had 20% lower odds of being screened (OR 0.8, 95% CI 0.7-0.9), compared with female patients, independent of patient age and provider characteristics (Table 2). Age was also associated independently with occurrence of screening, with older ages being associated in a dose-dependent manner: 18-year-olds had a more than 3-fold odds, compared with 12- to 13-year-olds. Examinations of provider characteristics (age, gender, tenure) suggested that male provider gender and older provider age may be associated positively with the occurrence of screening, but statistical tests could not exclude the role of chance in these associations. However, the provider-level factor from the multilevel modeling indicated significant contribution to the variance matrix (variance component 2.2; 95% CI 1.4-3.4), suggesting a possible individual provider effect; put another way, unmeasured characteristics of individual providers that could be related to screening decisions in the clinic could not be included in these models.

Table 2.

Odds of Tobacco Screening Among American Indian/Alaska Native Adolescent Patients (n = 4757).

	Odds Ratio	95% CI	P
Patient factors
Male gender	0.79	0.67-0.94	.007
Age (ref 12-13 years)
14-15 years	1.54	1.07-2.23	.019
16-17 years	1.67	1.21-2.32	.002
18 years	3.22	2.16-4.85	<.001
Provider factors
Male provider	1.63	0.68-3.94	.272
Provider age (ref 18-35 years)
36-55 years	1.19	0.54-2.46	.725
>55 years	1.48	0.58-3.78	.419
Provider tenure (ref <2 years)
2-5 years	1.17	0.45-3.49	.743
>5-10 years	1.30	0.44-3.86	.634
>10-15 years	1.62	0.38-6.89	.512
>16 years	0.46	0.11-1.93	.289
	Vara	95% CI
Provider ID	2.21	1.43-3.41

^aVariance component for the intercept of the random effects.

Similar regressions stratified by patient gender, indicated similar associations for patient age and provider characteristics, with little evidence of differences between genders (Table 3). Providers with increasing length of tenure may be increasingly less likely to screen for tobacco use, but this association was statistically significant only among male patients and providers with the most extreme tenure (>16 years); after accounting for multiple comparisons (α 0.05/3 tests) the significance was marginal.

Table 3.

Odds of Tobacco Screening Among American Indian/Alaska Native Adolescent Patients, Stratified by Patient Gender.

	Females (n = 2496)			Males (n = 2261)
	Odds Ratio	95% CI	P	Odds Ratio	(95% CI)	P
Patient factors
Age (ref 12-13 years)
14-15 years	1.67	1.05-2.64	.031	1.35	0.87-2.10	.182
16-17 years	1.93	1.25-2.98	.003	1.33	0.84-2.10	.223
18 years	4.21	2.69-6.58	<.001	2.03	1.19-3.47	.009
Provider factors
Male provider	1.83	0.73-4.57	.198	1.70	0.66-4.39	.276
Provider age (ref 18-35 years)
36-55 years	1.09	0.45-2.61	.847	1.32	0.59-2.97	.500
> 55 years	1.57	0.63-3.90	.340	1.57	0.53-4.67	.413
Provider tenure (ref <2 years)
2-5 years	1.68	0.58-4.90	.344	0.80	0.28-2.32	.685
>5-10 years	1.55	0.47-5.06	.468	1.03	0.32-3.27	.965
>10-15 years	2.07	0.44-9.64	.356	1.43	0.30-6.80	.654
>16 years	0.84	0.17-4.31	.837	0.19	0.04-0.85	.030
	Vara	95% CI		Vara	95% CI
Provider ID	2.23	1.42-3.48		2.43	1.51-3.93

^aVariance component for the intercept of the random effects.

Examinations of the occurrence of tobacco screening within specific combinations of patient and provider gender suggested that male providers had between 26% and 78% greater odds of screening compared with female providers, independent of other variables (Table 4). These statistically significant associations were consistent with the nonsignificant point estimates in the prior regression tables.

Table 4.

Odds of Tobacco Screening by Patient-Provider Dyads.

	Guideline-Based Screening (Crude Model)			Guideline-Based Screening (Adjusted Model)a
	Odds Ratio	95% CI	P	Odds Ratio	95% CI	P
Patient-provider dyad
Female provider, female patient (Referent)	1.00	(Ref)		1.00	(Ref)
Female provider, male patient	1.02	0.88-1.18	.786	0.95	0.82-1.10	.519
Male provider, male patient	1.52	1.29-1.79	<.001	1.26	1.05-1.50	.011
Male provider, female patient	1.78	1.45-2.17	<.001	1.43	1.15-1.77	.001

^aAdjusted for provider age (categorical), provider tenure (categorical), patient age (categorical).

Discussion

This study provided the first large-scale examination of patient and provider characteristics that influence clinical tobacco screening in AI/AN adolescents, allowing direct examination of the patient or provider characteristics. We found that only slightly more than one-third of all adolescent patients were screened for tobacco use in this sample; older patients, female patients, and male providers had the strongest positive associations with the occurrence of screening, although the male provider finding should be interpreted with caution due to limited statistical evidence. However, these data suggest that younger patients, male patients, and female providers may be missing from comprehensive clinic adherence to tobacco screening protocols. In addition, there was a high degree of individual provider contribution to variability in screening.

Previous research suggests that patient-provider gender concordant pairing results in better information gathering, communication, and therapeutic relationship within the medical encounter.^40,41 We found that male providers had higher odds for screening patients after accounting for patient gender, which conflicts with previous research. However, the location of the medical encounter may play a confounding role in these associations; rural adolescents were found to have lower screening rates than did their urban counterparts, and only 6 patient visits in the sample were with a male provider in a rural setting. It was not possible to examine these characteristics separately in the current study.

Primary care–based screening and referral for tobacco treatment among AI/AN adolescents is an integral step in reducing tobacco use among AI/AN people.⁴² The US Preventive Services Task Force recommends that screening and behavioral counseling for tobacco use occur for adults in primary care, but adolescents have received less policy attention.¹⁶ Data from the 2013 Alaska Youth Risk Behavior Survey indicate that 1 in 6 Alaska Native students initiated tobacco use before age 13 years, indicating the need to screen for tobacco use at a young age within the AI/AN population.⁴³ The low prevalence of tobacco screening observed in these data suggests that AI/AN adolescent tobacco users, in particular male AI/AN adolescents, have an unacceptably low likelihood of being identified, precluding clinical support for tobacco use cessation before progression to nicotine addiction.

Although this study was the first to examine tobacco screening among AI/AN adolescents, it was not without limitations. Family income variables were not available in our data set, and differences in health outcomes by socioeconomic factors may contribute to differences in tobacco use screening.⁴¹ Although racial concordance has been associated with constructs that improve medical interaction, such as increased cultural sensitivity,⁴⁴ provider race was not available in our dataset; analysis of the ways in which this variable may be related to screening is a question for future research. All patient-level data were recorded as part of routine health care and extracted from existing electronic medical records, meaning that misclassification with respect to the outcome may have contributed to overall study error. Some error is expected to be nondifferential and therefore primarily a limitation of statistical power. It is unknown how AI/AN adolescents interpret “tobacco use” in the screening questions, and may or may not include traditional tobacco use, electronic cigarettes, occasional tobacco use, or smokeless tobacco use. Finally, this study was restricted to active users of SCF services. Nearly 40% of Alaskan AI/ANs are not served by this health care system, and the extent to which findings generalize to other systems is not known.

Primary care–based screening and referral for tobacco treatment among AI/AN adolescents is an integral step in the reduction of tobacco use among AI/AN people, yet only 37% of AI/AN adolescents were screened for tobacco use in this study cohort. The low prevalence of tobacco screening observed in these data suggests that AI/AN adolescent tobacco users, in particular younger and male AI/AN adolescents, have an unacceptably low probability of being identified, precluding clinical support for tobacco use cessation before progression to nicotine addiction. Health system policy surrounding tobacco screening in AI/AN adolescent serving institutions should consider screening for tobacco use at every primary care visit to better address adolescent tobacco use. The odds ratios for screening increased with successive age groups. Higher screening odds were observed for male providers, irrespective of patient gender, which may comprise a potential target for provider education. Within this health system, adolescent tobacco screening occurs with a parent or guardian present which may impact provider decisions to offer screening and adolescent patient valid response to screening; future research that examines patient reporting on a sensitive topic is needed to determine how patient and parental attitudes toward tobacco use influence tobacco screening responses and provider adherence to the screening protocol in the presence of the parent or guardian. Future research focusing on patient and provider characteristics related to the occurrence of tobacco screening behavior is needed in order to develop understanding of the ways in which medical encounters with AI/AN adolescents may be better suited to identify tobacco use and thereby develop appropriate health interventions for tobacco use cessation. Clinical practice changes could include private screening of adolescents, changes in the wording of the screening question to include electronic cigarette and nonconventional tobacco products, and electronic reminders to screen during every adolescent primary care encounter.

Author Biographies

Vanessa Y. Hiratsuka, PhD MPH, (Navajo and Winnemum Wintu) is a senior researcher with the Research Department at Southcentral Foundation in Anchorage, Alaska. Her primary research interests are screening, detection, and management of chronic disease and behavior health issues within an American Indian/Alaska Native primary care setting and community-based participatory research with Alaska Native communities.

Astrid M. Suchy-Dicey, PhD, is an epidemiologist and biostatistician with the Center for Clinical and Epidemiological Research at the University of Washington in Seattle. Her research interests include etiology and molecular epidemiology of chronic and metabolic diseases related to aging in underserved and heavily burdened populations.

Eva M. Garroutte received a PhD in sociology from Princeton University in 1993 and is now a tenured research professor in the Department of Sociology at Boston College.  An enrolled citizen of the Cherokee Nation, Dr. Garroutte has a background of research and publication related to the study of racial-ethnic identity, religion, and American Indian health.  

Cathryn Booth-LaForce, PhD, is the Charles and Gerda Spence professor of Nursing in the department of Family & Child Nursing at the University of Washington. She is a developmental psychologist whose research interests include individual differences in social-emotional development from early infancy through late adolescence, using an ecobiodevelopmental framework; and early preventive intervention programs for American Indian and Alaska Native children and their caregivers.